| Summary: | 碩士 === 中華科技大學 === 電子工程研究所碩士班 === 99 === The independent component analysis (ICA) proposed by Pierre Comom in 1994, is one of the techniques to pick up the signal characteristics and to reduce the material relevant blind signal. In order to enhance its efficiency, Aapo Hyvarinen in 1997 proposed the fast independent component analysis(FastICA)which uses the kurtosis (kurtosis) fast fixed-point algorithm to separate signals, and can be applied
to the fields of biomedicine, speech recognition, pattern recognition, and so on.
This thesis proposed approaches to reduce the noise and clearly separate the primary signal. The discrete wavelet transform (DWT) is applied for noise reduction first. Afterwards, the signal is processed by Fast Independent Component Analysis approach to separate the primitive signal.
Selection law of the wavelet coefficient threshold value is applied for noise reduction. It achieves the goal of reducing noise by eliminating the noise which concentrate in the wavelet coefficients of high frequencies.
The architecture is designed by using the Verilog-HDL hardware description language. The result is simulated and verified by ModelSim. Finally it is implemented by the TSMC 0.18um technology.
|
|---|
